Professional Documents
Culture Documents
Distributed Hierarchical Game-Based Algorithm For Downlink Power Allocation in OFDMA Femtocell Networks PDF
Distributed Hierarchical Game-Based Algorithm For Downlink Power Allocation in OFDMA Femtocell Networks PDF
Computer Networks
journal homepage: www.elsevier.com/locate/comnet
a r t i c l e
i n f o
Article history:
Received 7 May 2015
Revised 15 September 2015
Accepted 10 November 2015
Available online 19 November 2015
Keywords:
Femtocell network
OFDMA
Game theory
Power allocation
a b s t r a c t
Femtocell is a promising technique for mobile operators to improve coverage and to provide
high-data rate services in a cost-ecient manner. In this study, we investigate the problem of
downlink power allocation in an orthogonal frequency division multiple access femtocell network. Femto-access points and macro-base stations in the network maximize their capacity
and compete with each other through power allocation. This competition is captured in the
framework of a Stackelberg game, and the Stackelberg equilibrium is introduced as the problem solution. We divide the strategy of MBSs into two substrategies (channel selection strategy
and power strategy) to improve the performance of MBSs. The two substrategies interact with
each other in the process of achieving Stackelberg equilibrium. The hierarchical iterative algorithm is proposed for MBSs and FAPs to achieve the maximum capacity game equilibrium in a
distributed manner. A dynamic step size mechanism and a method that copes with variational
player number are proposed to decrease the algorithm cost. Numerical results show that the
performance of the proposed algorithm is better than the performances of other algorithms.
2015 Elsevier B.V. All rights reserved.
1. Introduction
As a result of the widespread use of wireless devices and
the rapid diversication of wireless services, data trac in
wireless communication networks is explosively increasing.
More than 50% of voice services and 70% of data trac have
occurred indoors in recent years [1]. Thus, insucient indoor
coverage and the demand for high data rates are some of the
problems in existing cellular systems. Femtocell emerges as
one of the most promising technologies for improving indoor wireless service. Femto-access points (FAPs) are lowcost, low-power, and short-range base stations installed by
home users to improve indoor voice and data reception [2,3].
177
178
jk =
pkj
k k
i= j gi j pi
nkj
Ci =
kL
log2 1 +
pki
vki
skMi =
1,
0,
skNi =
i f pkNi > 0
otherwise
(active )
(inactive ).
up
up
Ci (Piup , Pi
, P low ) Ci (Piup , Pi
, P low ),
low
Ci Pilow , Pi
, P up
( (
179
) (
low
Ci Pilow , Pi
, P up
max
Pi
1+
log2
s.t. pki 0,
L
k k
j=i g ji p j
(6)
Existence of SE:
Proposition 1. [25] An NE exists in game G p,k = [M, pki , Cik ]. If
for all i M , then the following statements hold.
(1) pki is a non-empty convex and compact subset of some
Euclidean space RK .
(2) Cik ( pki ) is continuous and concave in pki .
Theorem 1. An NE exists in the subgame (upper and lower
games)G p,k = [M (N ), pki , Cik ]
Proof. The strategy space is dened as pki : 0 pki pmax
,
i
and it is a non-empty, convex, and compact subset of the Euclidean space RK .
Let
pki
Cik log2 1 + j=i gkji pkj +nki
1
=
= 2
.
pki
pki
ln (vki + pki )
Pj
(2)
(3)
(4)
(5)
+ nki
pki =pmax
i
Pj argmax C j : pkj 0,
)),
pki
k=1
(1)
(Pup , Plow ).
2Cik
1
= 2
< 0.
k 2
k
( pi )
ln (vi + pki )2
L
j M N.
pkj = pmax
,
j
k=1
(7)
We assume that all users can sense and differentiate a
useful signal from an interference signal. For description convenience, a single-leader and multiple-follower scenario is
formulated. To ease the notations, we let PM , SM , and Al replace PMi , SMi , and Ali , respectively.
We make the same assumption as in [22], i.e., subgames
converge to a unique and stable NE during the iterative waterlling (IW) process. The spectral radius of the channel gain
matrix should be less than unity for the stability and convergence of the IW process [27].
3.1. The solution of the lower game
Problem 1 : The lower game for the follower i is
PNi
L
log2 1 +
k=1
gkli pkM +
pkNi
k k
j=i g ji pN j
+ nkNi
(8)
k 0. Here, gk is the link
such that Lk=1 pkNi = pmax
and
p
Ni
Ni
li
gain from transmitter of the leader to the receiver of follower i on subchannel k. The noise-plus-interference term,
vkNi = gkli pkM + j=i gkji pkN j + nkNi , is treated as a constant by
the follower. The objective function is concave. Therefore, the
well-known IW algorithm can be the solution of the lower
game [27], i.e.,
PNi =
0
Ki vkNi
f or k
/ Afi
f or k A f i ,
(9)
max CM = max
PM
PM
L
k=1
log2
1+
pkM
k k
iN hi pNi
+ nkM
,
(10)
180
such that Lk=1 pkM = pmax
and pkM 0. Here, hki is the link
M
gain from the follower i transmitter to the leader receiver on
subchannel k. The solution is similar to the solution of the
lower game i.e.,
PM =
0
W vkM
f or k
/ Al
f or k Al ,
(11)
(12)
V = Gp + n,
(13)
n=
T
[n1N1 , . . . , nLN1 , n1N2 , . . . , nLN2 , n1NN , . . . , nLNN , n1M , . . . , nLM ] .
(14)
Matrix G can be interpreted as the gain matrix with gains
and zeros as entries. For example, when M = 1, N = 2, and
L = 2, the interference vector V is
1
0
vN
1
v2 0
N1
1 11
vN 2 g12
v 2 = g 21
N 2 12
1 11
vM h1
2
vM
h21
1
0
0
12
g12
22
g12
h12
1
h22
1
11
g21
21
g21
0
0
h11
2
h21
2
12
g21
22
g21
0
0
h12
2
h22
2
11
gl1
21
gl1
11
gl2
21
gl2
0
0
1 1
12
gl1
nN 1
pN 1
22 2
gl1
pN 1 n2N 1
1
12 1
gl2
p
n
N2 + N2 .
22 2
n2
gl2
p
N2 N2
0 p1M n1M
2
2
0
pM
nM
(15)
Matrix G is a partitioned matrix with zero diagonal blocks. Eq. (15) is the matrix expression of vkNi =
gkli pkM + j=i gkji pkN j + nkNi , f or Ni N and vkM = iN hki pkNi +
nkM , f or M M. Therefore, the interference of each user (in
both of the lower game and the upper game) on each subchannel can be found in V.
The game players do not need detailed knowledge of the
strategies of other players or the link gains. All they need is
the ability to know the power of useful signal and the interference. They can directly self-adjust their strategies by sensing the interference vki on their own subchannels. As a result, they can generate strategies in a distributed manner and
avoid learning high-dimensional information in Eq. (15). The
information of gains and the strategies of followers grow exponentially with the number of subchannels (or users) from
others.
The difference between the two subgames is the channel
selection. From the analysis of Section 2.2, we know that the
leader can abandon the bad subchannel voluntarily, even if
the subchannel is a positive prot for it. By contrast, given
that the followers cannot do that, they can only utilize the
leftover resource. If they can generate a positive prot from
the subchannel, then they take it or leave it.
Therefore, if both the strategy of the followers and the
channel selection strategy of the leader are constant, then we
k
CM
= log2 1 +
pk
vk
= log2 1 +
W vk
vk
= log2
W
vk
(16)
and
k
log2 1 + hpk Mpk +nk
CMk
iN i Ni
M
=
pkM
pkM
=
=
=
ln
1
k k
k
k
iN hi pNi + nM + pM
1
ln (vkM + pkM )
vkM
is the step size of the channel selection strategy searching, and we assume that = 1 to simplify the instruction.
The dynamic step size mechanism is explained in the ensuing
paragraphs.
Fig. 3 shows the detailed process of the hybrid strategy.
The left part is performed by the leader, and the right part is
performed by the followers. Initialize means the power allocation of SE without the strategy that the leader voluntarily
abandons a subchannel and T = 0. First, the leader estimates
the quality of all subchannels, and then obtains Q. Second,
the leader arrange Q in ascending order; thus, the leader can
know which subchannel quality is low. Third, the leader tentatively abandons subchannel that has the minimum quality. The leader then observes SE and estimates whether its
utility increases. If the utility increases, then the abandon
strategy is correct, and the leader creates more holes. Otherwise, the abandon strategy is incorrect. Hence, the algorithm is ended, and PM (T 1 ), PN (T 1 ) are the best solutions for the hybrid strategy in the Stackelberg game.
The proposed algorithm is distributed. Users do not need
to know the specic strategy of one another and the channel
gain with other transmitters. They generate a strategy by observing the interference and the useful signal they received.
Users can adaptively achieve the best strategy with the proposed algorithm.
1
2
ln W
(17)
Qk =
181
, k Al .
(18)
if (I ) (I 1 )
set = + Fo
else
set = max( Fo, 2 )
end if
end if
else
set = 1
the leader searches the best channel selection strategy
around SM (T 1 ) by the step size = 1
26:
the leader can reuse the highest-quality subchannel kL
kL
= 1) or abandon
in the abandoned subchannels (set SM
k1
= 0) tentatively
the lowest-quality subchannel k1 (set SM
27:
until achieve the maximum utility
28: end if
1:
2:
3:
4:
5:
6:
7:
8:
9:
10:
11:
12:
13:
14:
15:
16:
17:
18:
19:
20:
21:
22:
23:
24:
25:
182
Given that the proposed algorithm in this study is distributed, the players in the game adjust their strategies based
on sensory information (the interference and the useful signal). They do not require knowing the details of the interference source and the strategies of other players. Hence, they
can cope with the variational players easily. The mechanism
for the countermeasure of variational players is as follows:
If the leader senses that the variation in interference
exceeds a value (which is stated by the system), then it
will abandon the lowest-worth subchannel tentatively. If the
abandon strategy can improve its utility, then the leader
applies the abandon strategy until the maximum utility is
achieved, or it reuses the highest-worth subchannel that is
not occupied until the maximum utility is achieved.
The initial value of the original players should therefore
be the strategy in the previous stage. The power does not
need to be initialized to zero or an average value. Users can
adaptively adjust their strategy to achieve the maximum utility. The mechanism makes history information useful to decrease the strategy change frequency of users.
4.4. Convergence of the algorithm
Proposition 2. The lower game is convergent if the upper game
strategy is xed.
Proof. In the lower game, the IW algorithm is used to
achieve NE with xed power allocation of the leader. Thus,
in the process of achieving the lower game NE, the followers
183
Extension to multiple leaders. In the multiple-leader scenario, the interferences from the strategies of other leaders
are regarded as the background noise of each leader. The
leaders iterate their best response among themselves until
the upper subgame NE is reached. The followers use the waterlling algorithm to reach the lower subgame NE. Thus, the
process of searching the NE of the two subgames is iterated
until the entire process converges.
Scalability analysis of the algorithm. In the game, we offer the upper-tier users (MBS) a more exible strategy (hybrid strategy) to maintain high-level priority for users. Thus,
the algorithm can be used in other hierarchical networks.
The proposed algorithm is a fully distributed model. The algorithm can be easily extended to any other multiple-user
multiple-subchannel wireless networks, especially in highinterference conditions. Considering that we add the channel selection strategy based on the IW algorithm, the proposed algorithm performance cannot be lower than the IW
algorithm even in friendly interference conditions.
5. Numerical results
In this section, we present the simulation results to evaluate the performance of the proposed algorithm in comparison with the performances of the IW algorithm and the algorithm in [22].
A single-leader and two-follower scenario is considered
in the simulations. The amount of subchannels L = 100, and
the total power of a FAP, follower, is taken to be 20 dB bemax =
low the total power of MBS, leader. We assume that PM
30 dBm, and PNmax = 10 dBm. The noise power in each subchannel is WGN with a power of 120 dBm. Each subchannel is assumed to undergo independent and identically distributed Rayleigh fading. For each subchannel, the Rayleigh
random variable is modeled as the modulus of circularly
symmetrical complex Gaussian (CSCG) distributed. The average gain of a subchannel is controlled by changing the variance of the CSCG distribution. We set the iterative precision
105 . To verify the universal validity of the algorithm, we
base each plot on 100 random channel gain samples and observe the average performance, as shown in Table 1.
Table 1 shows the capacity (utility) of users with different algorithms. We are particularly interested in examining
the performance under high-interference conditions. The average direct channel gain is maintained to be unity, while the
average co-channel gain is kept to be 0.4 and 0.5. The second,
third, and fourth columns show the average capacity of the
leader, the followers, and the entire system, respectively.
From Table 1, the capacity of both the leader and the followers at SE is improved in contrast to NE. The capacity of
the leader is higher in a higher-gain condition. The leader
has more power than the follower, and the preponderance
is more obvious in a higher-gain condition. When the cochannel gain is higher, then the average co-channel gain is
0.5, and the algorithm in [22] makes the followers superior. The capacity of the followers is 97.3750, which is larger
than the capacity with the other algorithms. The leader creates signicant holes, thereby leading to the higher capacity of the followers, without increasing the capacity of the
leader. However, the capacity of the leader is two orders of
184
IW (NE)
[22] (SE)
This paper (SE)
Avg. co-ch. gain 0.5
IW (NE)
[22] (SE)
This paper (SE)
1538.2
1582.7
1694.6
Capacity of the leader
1551.3
1558.7
1710.1
23.2570
38.2557
42.4782
Capacity of the followers
20.3323
97.3750
38.8158
1561.46
1620.96
1737.08
Capacity of the system
1571.60
1656.08
1748.92
1.4
leader
followers
1.2
CDF
0.8
0.6
0.4
0.2
1.2
1.4
1.6
1.8
2
C /C
SE
2.2
2.4
2.6
2.8
NE
Fig. 4. Ratio of capacities at Stackelberg equilibrium using our algorithm and Nash equilibrium when average co-channel gain is 0.4.
Under different co-channel gains, Figs. 5 and 7 show the ratio of capacity at SE using our algorithm and the algorithm in
[22], respectively. When the average co-channel gain is 0.4,
approximately 70% of the ratios of the capacity of the followers obtained by our algorithm are higher than those obtained by the algorithm in [22]. When the average co-channel
gain is 0.5, the capacity of the followers obtained by the algorithm in [22] is higher than that obtained by our algorithm.
However, 96% of the ratios of the capacity of the leader (or
system capacity) obtained by our algorithm are higher than
that obtained by the algorithm in [22] under 0.4 average cochannel gain, and almost 100% of the ratios of the capacity
of the leader (or system capacity) obtained by our algorithm
are higher than that obtained by the algorithm in [22] under 0.5 average co-channel gain. The reason for this result is
that our algorithm focuses on preserving the priority of the
leader. Thus, in such a case, our algorithm shows a higher
performance in vast majority of cases.
We plot the iterative process in comparison with constant
step = 1 and dynamic step when the average co-channel
gain is 0.4 to show the dynamic spectral allocation process of
the proposed algorithm and the performance of the dynamic
step mechanism. The initial strategy of players is assigned
/L, pkNi =
with equal power on every subchannel, pkM = pmax
M
185
1.4
leader
followers
1.2
CDF
0.8
0.6
0.4
0.2
0
0.4
0.6
0.8
1.2
1.4
1.6
CSE/CSE[22]
1.8
2.2
Fig. 5. Ratio of capacities with our algorithm and the algorithm [22] when average co-channel gain is 0.4.
1.4
leader
followers
1.2
CDF
0.8
0.6
0.4
0.2
0
0.5
1.5
2
C
2.5
/C
SE
3.5
4.5
NE
Fig. 6. Ratio of capacities at Stackelberg equilibrium using our algorithm and Nash equilibrium when average co-channel gain is 0.5.
pmax
/L, f or k L. Two representative samples are shown as
Ni
follows.
Given that the capacity of the followers is not in the same
order of magnitude as that of the leader, the iterative process
curves of the followers are not given in gures. We show the
entire result in Table 2.
In Table 2, the rst two rows and the last two rows summarize the results in Samples 1 and 2, respectively.
Figs. 8 and 9 and Table 2 show that the proposed dynamic
step size mechanism largely improves the algorithm speed
186
1.4
leader
followers
1.2
CDF
0.8
0.6
0.4
0.2
0.2
0.4
0.6
0.8
CSE/CSE[22]
1.2
1.4
Fig. 7. Ratio of capacities with our algorithm and the algorithm [22] when average co-channel gain is 0.5.
Table 2
Performance of dynamic step size.
Algorithm
Leader change
strategy times
Leaders active
subchannels |Al |
Leader capacity
Follower change
strategy times
Followers active
subchannels |Af1 |, |Af2 |
Followers
capacity
=1
80
90
1672.4
308
15,10
Dynamic
45
88
1661.5
173
15,12
=1
73
91
1728.9
388
12,13
Dynamic
51
91
1728.6
292
12,12
24.5902
24.2341
28.2393
26.4867
28.4628
18.2036
25.5528
18.8200
187
Table 3
Effect of variational player number.
Player
|CCS|
Capacity before
follower 3 joins the
game
Capacity after
follower 3 joins the
game
Leader
Follower 1
Follower 2
Follower 3
4
5
4
-
1723.7
22.8610
22.4656
-
1641.60
20.0209
24.4961
20.8400
channel selection strategy based on the change in its capacity and generates a new power strategy until its capacity
achieves the maximum under the upper game equilibrium.
The hybrid strategy of leader is more exible and diverse
for maximizing its capacity. Followers use an IW to reach
the lower game equilibrium. The spectral holes created by
leaders are the main subchannels for followers. Thus, the interference of a leader from followers decreases to some extent. The two subgames are played iteratively to achieve SE.
A subchannel quality-estimating method, the dynamic step
size mechanism, and a method that copes with variational
player number are proposed to improve the performance of
the algorithm. The simulation results show the signicant capacity gains of macrocell and femtocell with a low cost.
Acknowledgment
The authors thank the nancial support of the Natural
Science Foundation of China under Grant nos. 61172095,
61571387 and 61473247.
Supplementary material
6. Conclusion
In this study, we propose a distributed algorithm to solve
the problem of downlink power allocation in OFDMA twotier femtocell networks. The objective of each station is to
maximize its capacity by power control with the total power
constraint. A hierarchical Stackelberg game is formulated.
In the upper game, we divide the strategy of a leader into
two substrategies (hybrid strategy), namely, the channel selection strategy and the power strategy. Leaders reach the
upper subgame equilibrium with their best hybrid strategies. In the process, each leader observes its capacity after
the power strategy is performed. A leader then adjusts its
188
[4] W.S. Jeon, J. Kim, D.G. Jeong, Downlink radio resource partitioning with
fractional frequency reuse in femtocell networks, IEEE Trans. Veh. Technol. 63 (2014) 308321.
[5] B. Wang, Y. Wu, K.R. Liu, Game theory for cognitive radio networks: An
overview, Comput. Netw. 54 (2010) 25372561.
[6] L. Song, D. Niyato, Z. Han, E. Hossain, Game-theoretic resource allocation methods for device-to-device communication, IEEE Wirel. Commun. 21 (2014) 136144.
[7] W. Wang, A. Kwasinski, D. Niyato, Z. Han, A survey on applications
of model-free strategy learning in cognitive wireless networks, arXiv
preprint arXiv:1504.03976 (2015).
[8] H.A. Shyllon, S. Mohan, A game theory-based distributed power control algorithm for femtocells, in: Proceedings of the 2014 IEEE International Conference Advance Networks Telecommunications Systems
(ANTS), 2014, pp. 16.
[9] J. Zhao, W. Zheng, X. Wen, X. Chu, H. Zhang, Z. Lu, Game theory
based energy-aware uplink resource allocation in OFDMA femtocell
networks, Int. J. Distrib. Sens. Netw. 2014 (2014) 18.
[10] H. Zhang, X. Chu, W. Ma, W. Zheng, X. Wen, Resource allocation with interference mitigation in OFDMA femtocells for co-channel deployment,
EURASIP J. Wirel. Commun. Netw. 2012 (2012) 19.
[11] S. Shen, T.M. Lok, Dynamic power allocation for downlink interference
management in a two-tier OFDMA network, IEEE Trans. Veh. Technol.
62 (2013) 41204125.
[12] Z. Liu, L. Hao, Y. Xia, X. Guan, Price bargaining based on the Stackelberg
game in two-tier orthogonal frequency division multiple access femtocell networks, IET Commun. 9 (2014) 133145.
[13] K. Zhu, E. Hossain, D. Niyato, Pricing, spectrum sharing, and service selection in two-tier small cell networks: A hierarchical dynamic game
approach, IEEE Trans Mob. Comput. 13 (2014) 18431856.
[14] Y. Xiao, G. Bi, D. Niyato, Distributed optimization for cognitive radio
networks using Stackelberg game, in: Proceedings of the 2010 IEEE International Conference Communication Systems (ICCS), 2010, pp. 77
81.
[15] W. Wang, A. Kwasinski, Z. Han, Power allocation with Stackelberg game
in femtocell networks: A self-learning approach, in: Proceedings of the
2014 IEEE Eleventh Annual International Conference on Sensing, Communication, and Networking (SECON), 2014, pp. 354362.
[16] X. Kang, R. Zhang, M. Motani, Price-based resource allocation for
spectrum-sharing femtocell networks: A Stackelberg game approach,
IEEE J. Sel. Areas Commun. 30 (2012) 538549.
[17] S. Sharma, D. Teneketzis, A game-theoretic approach to decentralized
optimal power allocation for cellular networks, Telecommun. Syst. 47
(2011) 6580.
[18] M. Bennis, M.L. Treust, S. Lasaulce, M. Debbah, J. Lilleberg, Spectrum
sharing games on the interference channel, in: Proceedings of the 2009
International Conference on Game Theory for Networks, 2009, pp. 515
522.
[19] Q. Han, K. Ma, X. Wang, X. Guan, J. Ma, Stackelberg game based interference management for two-tier femtocell networks, Wirel. Netw. 19
(2013) 16651677.
[20] Q. Han, B. Yang, X. Wang, K. Ma, C. Chen, X. Guan, Hierarchical-gamebased uplink power control in femtocell networks, IEEE Trans. Veh.
Technol. 63 (2014) 28192835.
[21] D.T. Ngo, S. Khakurel, T. Le-Ngoc, Joint subchannel assignment and
power allocation for OFDMA femtocell networks, IEEE Trans. Wirel.
Commun. 13 (2014) 342355.
[22] S. Guruacharya, D. Niyato, D.I. Kim, E. Hossain, Hierarchical competition for downlink power allocation in OFDMA femtocell networks, IEEE
Trans. Wirel. Commun. 12 (2013) 15431553.
[23] K. Son, S. Lee, Y. Yi, S. Chong, REFIM: A practical interference management in heterogeneous wireless access networks, IEEE J. Sel. Areas
Commun. 29 (2011) 12601272.
[24] H. Wang, Z. Ding, Power control and resource allocation for outage balancing in femtocell networks, IEEE Trans. Wirel. Commun. 14 (2015)
20432057.
[25] G. Owen, Game Theory, Academic, New York, NY, USA, 2005.
[26] K. Goebel, W.A. Kirk, Topics in Metric Fixed Point Theory, Cambridge
University Press, 1990.
[27] K.W. Shum, K.K. Leung, C.W. Sung, Convergence of iterative waterlling
algorithm for Gaussian interference channels, IEEE J. Sel. Areas Commun. 25 (2007) 10911100.
Zhixin Liu received his B.S., M.S., and Ph.D. degrees in Control Theory and Engineering from
Yanshan University, Qinhuangdao, China, in 2000,
2003, and 2006, respectively. He is currently a
professor with the Department of Automation,
Institute of Electrical Engineering, Yanshan University, China. He visited the University of Alberta, Edmonton, AB, Canada, in 2009. His current research interests include performance optimization and energy-ecient protocol design in
wireless sensor networks, wireless resource allocation in cognitive radio networks and femtocell
networks.
Xinping Guan received his Ph.D degree in Control
and Systems from the Harbin Institute of Technology, China. He joined the Department of Automation, Shanghai Jiao Tong University, China,
in 2007 where he is currently the Distinguished
University Professor, Director of Key Laboratory
of Systems Control and Information Processing,
Ministry of Education of China. Prof. Guan has
authored and/or coauthored two research monographs, more than 120 SCI indexed papers in IEEE
Transactions and other well-known international
journals and numerous conference papers. His
current research interests include wireless sensor
networks, ground-air communication of aircrafts, and cognitive radio networks and their applications in industry. As principle investigator, Prof. Guan
has nished/been working on many national key projects. He is the leader
of the prestigious Innovative Research Team of NSFC in 2012. He was one
of the rst prize winners of the University Natural Science Award from the
Ministry of Education of China in 2003 and 2007, respectively. He received
the IEEE Transaction on Fuzzy Systems Outstanding Paper Award in 2008
and also received the second prize of the National Natural Science Award of
China. Prof. Guan was appointed as Changjiang Scholar by the Ministry
of Education of China, and State-level Scholar of New Century Bai Qianwan Talent Program of China. He received the National Outstanding Youth
Award from the National Natural Science Foundation of China (NSFC). He
is also a Cyber Principle Investigator with the Cyber Joint Innovation Center
founded by Zhejiang University, Tsinghua University, and Shanghai Jiao Tong
University. Prof. Guan is a senior member of IEEE. He is currently a committee member of the Chinese Automation Association Council and Chinese
Articial Intelligence Association Council. He serves as an Associate Editor
for IEEE Transaction on System, Man and Cybernetics-C, as Editorial Board
Committee Member for the several Chinese journals, and an International
Technical Committee Member for a lot of conferences. Wirel.